2011 - Cooperative Institute for Research in Environmental Sciences ...

The SEAESRT (Space, Environmental Expert System Real
Time) algorithm to specify satellite anomaly hazards was
developed and successfully tested before it was provided
to the Space Weather Prediction Center (SWPC). It has not
yet been put into operational test mode by SWPC because
of other priorities. The National Geophysical Data Center
(NGDC) may implement it into their system as a postanalysis
tool.
Research, development and testing of five algorithms for
Phase 2 of the GOES-R development project were completed
and delivered to SWPC. These algorithms will be
part of the suite of Level 2+ processing of GOES-R data
to provide the forecast center with specification, analysis
and forecast tools for the space environment. They include
the magnetometer analysis tool that indicates when a
magnetopause crossing has been detected at geostationary
altitude by the GOES-R satellite. Two more of the algorithms
were developed to use the energetic particle data to
determine temperature and density moments and indicate
spacecraft charging levels. The final two algorithms
analyze the solar ultra-violet imaging data to produce
thematic maps and coronal hole images of the sun.
Planning and requirements gathering for Phase 3 of the
algorithms development project are underway.
Product: Loto’aniu, TM, and HJ Singer (2011), The GOES-
R magnetopause crossing algorithm theoretical basis document,
NOAA/NESDIS Technical Publication.
Loto’aniu, TM, L Mayer, and M Berguson (2011), The
GOES-R magnetopause crossing algorithm implementation
and users’ guide, NOAA/NESDIS Technical Publication.
Loto’aniu, TM, L Mayer, and M Berguson (2011), The
GOES-R delivery 2 test plan and results, NOAA/NESDIS
Technical Publication.
Rodriguez, J (2011), GOES-R SEISS density and temperature
moments and level of spacecraft charging algorithm
theoretical basis document, NOAA/NEDIS Technical Document.
Rodriguez, J (2011), GOES-R SEISS density and temperature
moments and level of spacecraft charging algorithm
implementation and users’ guide, NOAA/NEDIS Technical
Document.
Rodriguez, J (2011,) GOES-R SEISS density and temperature
moments and level of spacecraft charging algorithm
test plan and results, NOAA/NEDIS Technical Document.
Rigler, EJ, and SM Hill (2011), SUVI thematic maps algorithm
theoretical basis document, NOAA/NEDIS Technical
Document.
GSD-07 High Performance
Computing Systems
FEDERAL LEAD: SCOTT NAHMAN
CIRES LEAD: CRAIG TIERNEY
NOAA Goals 2: Climate
Project Goal: Provide systems research support for high-performance
computing (HPC) efforts and assistance to the user
community; provide HPC Systems communications equipment
and software research; and provide research support for highperformance
file systems.
92 CIRES Annual Report 2011
Milestone 1. Conduct technical study of latest hardware
architectures to support future NOAA procurements.
NOAA has increased investment into new high-performance
computing (HPC) systems over the last year for
supporting both weather and hurricane modeling. While
investment has increased, the research needs continue to
outstrip resources. CIRES supported NOAA in the identification
and acquisition of two major HPC platforms. Our
goal in each case was to evaluate existing technologies and
lend our own expertise to the architecture and implementation
of these systems to maximize NOAA’s investment. The
first system was delivered at the end of summer 2010. This
system, tJet, was a three-times performance increase over
the previous system. This made the Earth System Research
Laboratory (ESRL) the home to the largest NOAA-managed
HPC resources.
The second system procured is a system whose purpose
is to support weather and other modeling efforts for the
entire NOAA program. The system is to be located in Fairmont,
W. Va. CIRES provided assistance in the gathering
of requirements, and made architectural recommendations
to NOAA to help them maximize their investment. This
system shall be operational by the end of 2011.
Milestone 2. Investigate tools to automate the use of
Graphical Processor Units (GPU) co-processors within existing
GSD codes.
While the use of HPC continues to increase in importance
and expand to other disciplines, the users of this technology
are starting to run into a performance wall. While the
number of cores per central processing unit (CPU) socket
continues to increase, the performance of individual cores
is not increasing. Also, the power consumed on these
systems is becoming a significant portion of the total cost of
ownership of HPC systems. NOAA’s Global Systems Division
(GSD) is looking at new hardware technologies that
can provide significant performance improvements while
decreasing power consumption. Graphical Processor Units
(GPUs) are one technology that appears to have the ability
to meet this need, but effective use of these devices requires
significant changes in programming models.
We have been supporting NOAA in the investigation of
commercial and open-source tools that can help automate
conversion of existing code to efficient code on the GPUs.
Several vendors (Portland Group, Intel, Cray, CAPS) are
taking different approaches to this problem. While most
of the technologies are still immature, they are all showing
promise. Comparisons of code generated by vendors’ tools
to handwritten code show tools are between 50 percent and
90 percent as efficient as handwritten code. Much work is
still needed to assist the vendors with their programming
tools to increase relative performance as close to 100 percent
as possible.
Milestone 3. Support investigations of large, core-count
model scalability in heterogeneous computing environments.
Due to time restraints and changes in NOAA priorities for
the high-performance computing acquisition research area,
we were not able to address this Milestone.